Forecasting of transportation-related energy demand and CO2 emissions in Turkey with different machine learning algorithms

Adverse impacts of the transportation sector on not only air quality but also economic growth of a country are nowadays well-noticed, particularly by developing countries. Today, the transportation sector is powered by burning the fossil-based fuels at more than 99% and approximately 6.5 million deaths annually occur due to air-pollution-related diseases worldwide. Therefore, knowledge of both energy demand and CO2 emission of a country is a very significant issue in order to revise its future energy investments and policies. In this framework, three machine learning algorithms (deep learning (DL), support vector machine (SVM), and artificial neural network (ANN)) are used to forecast the transportation-based-CO2 emission and energy demand in Turkey. The gross domestic product per capita (GDP), population, vehicle kilometer, and year are used as input parameters in the study. It is noticed that there is a very high correlation among year, economic indicators, population, vehicle kilometer, transportation-based energy demand, and CO2 emissions. To present a better comparison, the results of these algorithms are discussed with six frequently used statistical metrics (R-2, RMSE, MAPE, MBE, rRMSE, and MABE). For all machine learning algorithms, R-2 values are varying between 0.8639 and 0.9235, and RMSE is smaller than 5 x 10(6) tons for CO2 emission and 2 Mtoe for energy demand. According to the classifications in the literature, the forecast results are generally categorized as excellent for rRMSE metric (<10%), and high prediction accuracy for MAPE metric (